3 #include "repository.h"
11 #include "cache-tree.h"
12 #include "tree-walk.h"
14 #include "parse-options.h"
17 #include "streaming.h"
20 #include "object-store.h"
21 #include "run-command.h"
24 #define REACHABLE 0x0001
26 #define HAS_OBJ 0x0004
27 /* This flag is set if something points to this object. */
32 static int show_unreachable;
33 static int include_reflogs = 1;
34 static int check_full = 1;
35 static int connectivity_only;
36 static int check_strict;
37 static int keep_cache_objects;
38 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
39 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
40 static int errors_found;
41 static int write_lost_and_found;
43 static int show_progress = -1;
44 static int show_dangling = 1;
45 static int name_objects;
46 #define ERROR_OBJECT 01
47 #define ERROR_REACHABLE 02
49 #define ERROR_REFS 010
50 #define ERROR_COMMIT_GRAPH 020
52 static const char *describe_object(struct object *obj)
54 static struct strbuf bufs[] = {
55 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
62 name = lookup_decoration(fsck_walk_options.object_names, obj);
65 b = (b + 1) % ARRAY_SIZE(bufs);
67 strbuf_addstr(buf, oid_to_hex(&obj->oid));
69 strbuf_addf(buf, " (%s)", name);
74 static const char *printable_type(struct object *obj)
78 if (obj->type == OBJ_NONE) {
79 enum object_type type = oid_object_info(the_repository,
82 object_as_type(the_repository, obj, type, 0);
85 ret = type_name(obj->type);
92 static int fsck_config(const char *var, const char *value, void *cb)
94 if (strcmp(var, "fsck.skiplist") == 0) {
96 struct strbuf sb = STRBUF_INIT;
98 if (git_config_pathname(&path, var, value))
100 strbuf_addf(&sb, "skiplist=%s", path);
102 fsck_set_msg_types(&fsck_obj_options, sb.buf);
107 if (skip_prefix(var, "fsck.", &var)) {
108 fsck_set_msg_type(&fsck_obj_options, var, value);
112 return git_default_config(var, value, cb);
115 static int objerror(struct object *obj, const char *err)
117 errors_found |= ERROR_OBJECT;
118 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
119 fprintf_ln(stderr, _("error in %s %s: %s"),
120 printable_type(obj), describe_object(obj), err);
124 static int fsck_error_func(struct fsck_options *o,
125 struct object *obj, int type, const char *message)
129 /* TRANSLATORS: e.g. warning in tree 01bfda: <more explanation> */
130 fprintf_ln(stderr, _("warning in %s %s: %s"),
131 printable_type(obj), describe_object(obj), message);
134 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
135 fprintf_ln(stderr, _("error in %s %s: %s"),
136 printable_type(obj), describe_object(obj), message);
139 BUG("%d (FSCK_IGNORE?) should never trigger this callback", type);
143 static struct object_array pending;
145 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
147 struct object *parent = data;
150 * The only case data is NULL or type is OBJ_ANY is when
151 * mark_object_reachable() calls us. All the callers of
152 * that function has non-NULL obj hence ...
155 /* ... these references to parent->fld are safe here */
156 printf_ln(_("broken link from %7s %s"),
157 printable_type(parent), describe_object(parent));
158 printf_ln(_("broken link from %7s %s"),
159 (type == OBJ_ANY ? _("unknown") : type_name(type)),
161 errors_found |= ERROR_REACHABLE;
165 if (type != OBJ_ANY && obj->type != type)
166 /* ... and the reference to parent is safe here */
167 objerror(parent, _("wrong object type in link"));
169 if (obj->flags & REACHABLE)
171 obj->flags |= REACHABLE;
173 if (is_promisor_object(&obj->oid))
175 * Further recursion does not need to be performed on this
176 * object since it is a promisor object (so it does not need to
177 * be added to "pending").
181 if (!(obj->flags & HAS_OBJ)) {
182 if (parent && !has_object_file(&obj->oid)) {
183 printf_ln(_("broken link from %7s %s\n"
185 printable_type(parent),
186 describe_object(parent),
188 describe_object(obj));
189 errors_found |= ERROR_REACHABLE;
194 add_object_array(obj, NULL, &pending);
198 static void mark_object_reachable(struct object *obj)
200 mark_object(obj, OBJ_ANY, NULL, NULL);
203 static int traverse_one_object(struct object *obj)
205 int result = fsck_walk(obj, obj, &fsck_walk_options);
207 if (obj->type == OBJ_TREE) {
208 struct tree *tree = (struct tree *)obj;
209 free_tree_buffer(tree);
214 static int traverse_reachable(void)
216 struct progress *progress = NULL;
220 progress = start_delayed_progress(_("Checking connectivity"), 0);
222 result |= traverse_one_object(object_array_pop(&pending));
223 display_progress(progress, ++nr);
225 stop_progress(&progress);
229 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
238 * Check a single reachable object
240 static void check_reachable_object(struct object *obj)
243 * We obviously want the object to be parsed,
244 * except if it was in a pack-file and we didn't
247 if (!(obj->flags & HAS_OBJ)) {
248 if (is_promisor_object(&obj->oid))
250 if (has_object_pack(&obj->oid))
251 return; /* it is in pack - forget about it */
252 printf_ln(_("missing %s %s"), printable_type(obj),
253 describe_object(obj));
254 errors_found |= ERROR_REACHABLE;
260 * Check a single unreachable object
262 static void check_unreachable_object(struct object *obj)
265 * Missing unreachable object? Ignore it. It's not like
266 * we miss it (since it can't be reached), nor do we want
267 * to complain about it being unreachable (since it does
270 if (!(obj->flags & HAS_OBJ))
274 * Unreachable object that exists? Show it if asked to,
275 * since this is something that is prunable.
277 if (show_unreachable) {
278 printf_ln(_("unreachable %s %s"), printable_type(obj),
279 describe_object(obj));
284 * "!USED" means that nothing at all points to it, including
285 * other unreachable objects. In other words, it's the "tip"
286 * of some set of unreachable objects, usually a commit that
289 * Such starting points are more interesting than some random
290 * set of unreachable objects, so we show them even if the user
291 * hasn't asked for _all_ unreachable objects. If you have
292 * deleted a branch by mistake, this is a prime candidate to
293 * start looking at, for example.
295 if (!(obj->flags & USED)) {
297 printf_ln(_("dangling %s %s"), printable_type(obj),
298 describe_object(obj));
299 if (write_lost_and_found) {
300 char *filename = git_pathdup("lost-found/%s/%s",
301 obj->type == OBJ_COMMIT ? "commit" : "other",
302 describe_object(obj));
305 if (safe_create_leading_directories_const(filename)) {
306 error(_("could not create lost-found"));
310 f = xfopen(filename, "w");
311 if (obj->type == OBJ_BLOB) {
312 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
313 die_errno(_("could not write '%s'"), filename);
315 fprintf(f, "%s\n", describe_object(obj));
317 die_errno(_("could not finish '%s'"),
325 * Otherwise? It's there, it's unreachable, and some other unreachable
326 * object points to it. Ignore it - it's not interesting, and we showed
327 * all the interesting cases above.
331 static void check_object(struct object *obj)
334 fprintf_ln(stderr, _("Checking %s"), describe_object(obj));
336 if (obj->flags & REACHABLE)
337 check_reachable_object(obj);
339 check_unreachable_object(obj);
342 static void check_connectivity(void)
346 /* Traverse the pending reachable objects */
347 traverse_reachable();
349 /* Look up all the requirements, warn about missing objects.. */
350 max = get_max_object_index();
352 fprintf_ln(stderr, _("Checking connectivity (%d objects)"), max);
354 for (i = 0; i < max; i++) {
355 struct object *obj = get_indexed_object(i);
362 static int fsck_obj(struct object *obj, void *buffer, unsigned long size)
366 if (obj->flags & SEEN)
371 fprintf_ln(stderr, _("Checking %s %s"),
372 printable_type(obj), describe_object(obj));
374 if (fsck_walk(obj, NULL, &fsck_obj_options))
375 objerror(obj, _("broken links"));
376 err = fsck_object(obj, buffer, size, &fsck_obj_options);
380 if (obj->type == OBJ_COMMIT) {
381 struct commit *commit = (struct commit *) obj;
383 if (!commit->parents && show_root)
384 printf_ln(_("root %s"),
385 describe_object(&commit->object));
388 if (obj->type == OBJ_TAG) {
389 struct tag *tag = (struct tag *) obj;
391 if (show_tags && tag->tagged) {
392 printf_ln(_("tagged %s %s (%s) in %s"),
393 printable_type(tag->tagged),
394 describe_object(tag->tagged),
396 describe_object(&tag->object));
401 if (obj->type == OBJ_TREE)
402 free_tree_buffer((struct tree *)obj);
403 if (obj->type == OBJ_COMMIT)
404 free_commit_buffer((struct commit *)obj);
408 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
409 unsigned long size, void *buffer, int *eaten)
412 * Note, buffer may be NULL if type is OBJ_BLOB. See
413 * verify_packfile(), data_valid variable for details.
416 obj = parse_object_buffer(the_repository, oid, type, size, buffer,
419 errors_found |= ERROR_OBJECT;
420 return error(_("%s: object corrupt or missing"),
423 obj->flags &= ~(REACHABLE | SEEN);
424 obj->flags |= HAS_OBJ;
425 return fsck_obj(obj, buffer, size);
428 static int default_refs;
430 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
431 timestamp_t timestamp)
435 if (!is_null_oid(oid)) {
436 obj = lookup_object(the_repository, oid->hash);
437 if (obj && (obj->flags & HAS_OBJ)) {
438 if (timestamp && name_objects)
439 add_decoration(fsck_walk_options.object_names,
441 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
443 mark_object_reachable(obj);
444 } else if (!is_promisor_object(oid)) {
445 error(_("%s: invalid reflog entry %s"),
446 refname, oid_to_hex(oid));
447 errors_found |= ERROR_REACHABLE;
452 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
453 const char *email, timestamp_t timestamp, int tz,
454 const char *message, void *cb_data)
456 const char *refname = cb_data;
459 fprintf_ln(stderr, _("Checking reflog %s->%s"),
460 oid_to_hex(ooid), oid_to_hex(noid));
462 fsck_handle_reflog_oid(refname, ooid, 0);
463 fsck_handle_reflog_oid(refname, noid, timestamp);
467 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
468 int flag, void *cb_data)
470 struct strbuf refname = STRBUF_INIT;
472 strbuf_worktree_ref(cb_data, &refname, logname);
473 for_each_reflog_ent(refname.buf, fsck_handle_reflog_ent, refname.buf);
474 strbuf_release(&refname);
478 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
479 int flag, void *cb_data)
483 obj = parse_object(the_repository, oid);
485 if (is_promisor_object(oid)) {
487 * Increment default_refs anyway, because this is a
493 error(_("%s: invalid sha1 pointer %s"),
494 refname, oid_to_hex(oid));
495 errors_found |= ERROR_REACHABLE;
496 /* We'll continue with the rest despite the error.. */
499 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
500 error(_("%s: not a commit"), refname);
501 errors_found |= ERROR_REFS;
506 add_decoration(fsck_walk_options.object_names,
507 obj, xstrdup(refname));
508 mark_object_reachable(obj);
513 static int fsck_head_link(const char *head_ref_name,
514 const char **head_points_at,
515 struct object_id *head_oid);
517 static void get_default_heads(void)
519 struct worktree **worktrees, **p;
520 const char *head_points_at;
521 struct object_id head_oid;
523 for_each_rawref(fsck_handle_ref, NULL);
525 worktrees = get_worktrees(0);
526 for (p = worktrees; *p; p++) {
527 struct worktree *wt = *p;
528 struct strbuf ref = STRBUF_INIT;
530 strbuf_worktree_ref(wt, &ref, "HEAD");
531 fsck_head_link(ref.buf, &head_points_at, &head_oid);
532 if (head_points_at && !is_null_oid(&head_oid))
533 fsck_handle_ref(ref.buf, &head_oid, 0, NULL);
534 strbuf_release(&ref);
537 refs_for_each_reflog(get_worktree_ref_store(wt),
538 fsck_handle_reflog, wt);
540 free_worktrees(worktrees);
543 * Not having any default heads isn't really fatal, but
544 * it does mean that "--unreachable" no longer makes any
545 * sense (since in this case everything will obviously
546 * be unreachable by definition.
548 * Showing dangling objects is valid, though (as those
549 * dangling objects are likely lost heads).
551 * So we just print a warning about it, and clear the
552 * "show_unreachable" flag.
555 fprintf_ln(stderr, _("notice: No default references"));
556 show_unreachable = 0;
560 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
563 enum object_type type;
568 if (read_loose_object(path, oid, &type, &size, &contents) < 0) {
569 errors_found |= ERROR_OBJECT;
570 error(_("%s: object corrupt or missing: %s"),
571 oid_to_hex(oid), path);
572 return 0; /* keep checking other objects */
575 if (!contents && type != OBJ_BLOB)
576 BUG("read_loose_object streamed a non-blob");
578 obj = parse_object_buffer(the_repository, oid, type, size,
582 errors_found |= ERROR_OBJECT;
583 error(_("%s: object could not be parsed: %s"),
584 oid_to_hex(oid), path);
587 return 0; /* keep checking other objects */
590 obj->flags &= ~(REACHABLE | SEEN);
591 obj->flags |= HAS_OBJ;
592 if (fsck_obj(obj, contents, size))
593 errors_found |= ERROR_OBJECT;
597 return 0; /* keep checking other objects, even if we saw an error */
600 static int fsck_cruft(const char *basename, const char *path, void *data)
602 if (!starts_with(basename, "tmp_obj_"))
603 fprintf_ln(stderr, _("bad sha1 file: %s"), path);
607 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
609 display_progress(progress, nr + 1);
613 static void fsck_object_dir(const char *path)
615 struct progress *progress = NULL;
618 fprintf_ln(stderr, _("Checking object directory"));
621 progress = start_progress(_("Checking object directories"), 256);
623 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
625 display_progress(progress, 256);
626 stop_progress(&progress);
629 static int fsck_head_link(const char *head_ref_name,
630 const char **head_points_at,
631 struct object_id *head_oid)
633 int null_is_error = 0;
636 fprintf_ln(stderr, _("Checking %s link"), head_ref_name);
638 *head_points_at = resolve_ref_unsafe(head_ref_name, 0, head_oid, NULL);
639 if (!*head_points_at) {
640 errors_found |= ERROR_REFS;
641 return error(_("invalid %s"), head_ref_name);
643 if (!strcmp(*head_points_at, head_ref_name))
646 else if (!starts_with(*head_points_at, "refs/heads/")) {
647 errors_found |= ERROR_REFS;
648 return error(_("%s points to something strange (%s)"),
649 head_ref_name, *head_points_at);
651 if (is_null_oid(head_oid)) {
653 errors_found |= ERROR_REFS;
654 return error(_("%s: detached HEAD points at nothing"),
658 _("notice: %s points to an unborn branch (%s)"),
659 head_ref_name, *head_points_at + 11);
664 static int fsck_cache_tree(struct cache_tree *it)
670 fprintf_ln(stderr, _("Checking cache tree"));
672 if (0 <= it->entry_count) {
673 struct object *obj = parse_object(the_repository, &it->oid);
675 error(_("%s: invalid sha1 pointer in cache-tree"),
676 oid_to_hex(&it->oid));
677 errors_found |= ERROR_REFS;
682 add_decoration(fsck_walk_options.object_names,
684 mark_object_reachable(obj);
685 if (obj->type != OBJ_TREE)
686 err |= objerror(obj, _("non-tree in cache-tree"));
688 for (i = 0; i < it->subtree_nr; i++)
689 err |= fsck_cache_tree(it->down[i]->cache_tree);
693 static void mark_object_for_connectivity(const struct object_id *oid)
695 struct object *obj = lookup_unknown_object(oid->hash);
696 obj->flags |= HAS_OBJ;
699 static int mark_loose_for_connectivity(const struct object_id *oid,
703 mark_object_for_connectivity(oid);
707 static int mark_packed_for_connectivity(const struct object_id *oid,
708 struct packed_git *pack,
712 mark_object_for_connectivity(oid);
716 static char const * const fsck_usage[] = {
717 N_("git fsck [<options>] [<object>...]"),
721 static struct option fsck_opts[] = {
722 OPT__VERBOSE(&verbose, N_("be verbose")),
723 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
724 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
725 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
726 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
727 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
728 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
729 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
730 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
731 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
732 OPT_BOOL(0, "lost-found", &write_lost_and_found,
733 N_("write dangling objects in .git/lost-found")),
734 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
735 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
739 int cmd_fsck(int argc, const char **argv, const char *prefix)
742 struct object_directory *odb;
744 /* fsck knows how to handle missing promisor objects */
745 fetch_if_missing = 0;
748 read_replace_refs = 0;
750 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
752 fsck_walk_options.walk = mark_object;
753 fsck_obj_options.walk = mark_used;
754 fsck_obj_options.error_func = fsck_error_func;
756 fsck_obj_options.strict = 1;
758 if (show_progress == -1)
759 show_progress = isatty(2);
763 if (write_lost_and_found) {
769 fsck_walk_options.object_names =
770 xcalloc(1, sizeof(struct decoration));
772 git_config(fsck_config, NULL);
774 if (connectivity_only) {
775 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
776 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
778 prepare_alt_odb(the_repository);
779 for (odb = the_repository->objects->odb; odb; odb = odb->next)
780 fsck_object_dir(odb->path);
783 struct packed_git *p;
784 uint32_t total = 0, count = 0;
785 struct progress *progress = NULL;
788 for (p = get_all_packs(the_repository); p;
790 if (open_pack_index(p))
792 total += p->num_objects;
795 progress = start_progress(_("Checking objects"), total);
797 for (p = get_all_packs(the_repository); p;
799 /* verify gives error messages itself */
800 if (verify_pack(the_repository,
803 errors_found |= ERROR_PACK;
804 count += p->num_objects;
806 stop_progress(&progress);
809 if (fsck_finish(&fsck_obj_options))
810 errors_found |= ERROR_OBJECT;
813 for (i = 0; i < argc; i++) {
814 const char *arg = argv[i];
815 struct object_id oid;
816 if (!get_oid(arg, &oid)) {
817 struct object *obj = lookup_object(the_repository,
820 if (!obj || !(obj->flags & HAS_OBJ)) {
821 if (is_promisor_object(&oid))
823 error(_("%s: object missing"), oid_to_hex(&oid));
824 errors_found |= ERROR_OBJECT;
830 add_decoration(fsck_walk_options.object_names,
832 mark_object_reachable(obj);
835 error(_("invalid parameter: expected sha1, got '%s'"), arg);
836 errors_found |= ERROR_OBJECT;
840 * If we've not been given any explicit head information, do the
841 * default ones from .git/refs. We also consider the index file
842 * in this case (ie this implies --cache).
846 keep_cache_objects = 1;
849 if (keep_cache_objects) {
850 verify_index_checksum = 1;
853 for (i = 0; i < active_nr; i++) {
858 mode = active_cache[i]->ce_mode;
859 if (S_ISGITLINK(mode))
861 blob = lookup_blob(the_repository,
862 &active_cache[i]->oid);
868 add_decoration(fsck_walk_options.object_names,
870 xstrfmt(":%s", active_cache[i]->name));
871 mark_object_reachable(obj);
873 if (active_cache_tree)
874 fsck_cache_tree(active_cache_tree);
877 check_connectivity();
879 if (!git_config_get_bool("core.commitgraph", &i) && i) {
880 struct child_process commit_graph_verify = CHILD_PROCESS_INIT;
881 const char *verify_argv[] = { "commit-graph", "verify", NULL, NULL, NULL };
883 prepare_alt_odb(the_repository);
884 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
885 child_process_init(&commit_graph_verify);
886 commit_graph_verify.argv = verify_argv;
887 commit_graph_verify.git_cmd = 1;
888 verify_argv[2] = "--object-dir";
889 verify_argv[3] = odb->path;
890 if (run_command(&commit_graph_verify))
891 errors_found |= ERROR_COMMIT_GRAPH;
895 if (!git_config_get_bool("core.multipackindex", &i) && i) {
896 struct child_process midx_verify = CHILD_PROCESS_INIT;
897 const char *midx_argv[] = { "multi-pack-index", "verify", NULL, NULL, NULL };
899 prepare_alt_odb(the_repository);
900 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
901 child_process_init(&midx_verify);
902 midx_verify.argv = midx_argv;
903 midx_verify.git_cmd = 1;
904 midx_argv[2] = "--object-dir";
905 midx_argv[3] = odb->path;
906 if (run_command(&midx_verify))
907 errors_found |= ERROR_COMMIT_GRAPH;